This repository collects small experiments and utilities for characterizing the spatial organization of cells in tissue slices. The code focuses on comparing ground-truth and predicted cell layouts with metrics drawn from spatial statistics, graph spectra, graph portraits, and continuous energy landscapes.
Utilities for loading cell-level CSV data into AnnData objects and measuring
gene-expression deviations across spatial grids.
Ripley's K, Besag's L, and pair-correlation functions for quantifying whether cells of a chosen type are clustered, dispersed, or approximately random across the tissue plane.
Graph Laplacian tools for building spatial cell graphs, estimating connectivity, computing spectral gaps, and measuring Von Neumann entropy from graph spectra.
Continuous landscape builders and Cahn-Hilliard-style energy measurements for turning discrete cell coordinates into fields that can be compared by spatial phase structure.
To measure a single cell type cohesion we can compute the energy of the tissue landscape considering that cells are from a phase and the empty space from the opposite phase.
To measure how mixed different cell types are we can build landscapes from 2 different cell types using Voronoi like uilt diagrams and the compute the Cahn-Hilliard energy of these landscapes
Notebook experiments using graph portrait comparisons to evaluate similarities between cell-type-specific spatial graphs. This was made using Graph-Compass library
